Issue 178 Item 10 Automated Screening of Color Photos May Identify Diabetic Reti

An automated system of color retinal image analysis represents a potential new screening method for exudative diabetic retinopathy. Good news, according to a report in the British Journal of Ophthalmology for October.

Dr. Alireza Osareh at the University of Bristol and colleagues used 75 color retinal images (25 normal, 50 abnormal) to develop their model. The first step involved color normalization and local contrast enhancement against a reference image.

The authors then employed an artificial neural network classifier to analyze color regions at pixel resolution, differentiated on the basis of such features as color, size and texture. Clinicians labeled each region as an exudate or non-exudate.

When tested against another group of images (27 normal, 40 abnormal), the researchers observed 95% sensitivity in identifying affected retinas and 88.9% specificity.

According to Dr. Osareh’s group, their system "could be used to evaluate digital retinal images obtained in screening programmes for diabetic retinopathy and used by non-experts to indicate which patients require referral to an ophthalmologist for further investigation and treatment." Br J Ophthalmol 2003;2003;87:1220-1223.